- There is no need to branch on the rethrow variable to determine if
we need to call objc_exception_try_exit. We know whether an
exception was thrown, so just branch to a different target if we
want to skip the try_exit.
This is a slight semantic departure from gcc, but only for throwing
nil, which is undefined (and for which gcc emits broken code).
- Also fixes a bug in current code which was calling try_exit too
many times when an exception was uncaught (but there were some
handlers).
Fix bug introduced in prev. commit, the type of the @catch parameter
was uninitialized.
llvm-svn: 56754
when there are no @catch blocks.
Also, eliminated unneeded alloca for the rethrow object.
Also, avoid generating code to rethrow exception if the exception will
trivially always be caught.
llvm-svn: 56751
- return attributes - inreg, zext and sext
- parameter attributes
- function attributes - nounwind, readonly, readnone, noreturn
Return attributes use 0 as the index.
Function attributes use ~0U as the index.
llvm-svn: 56705
to whether the fileid is a 'extern c system header' in addition to whether it
is a system header, most of this is spreading plumbing around. Once we have that,
PPLexerChange bases its "file enter/exit" notifications to PPCallbacks to
base the system header state on FileIDInfo instead of HeaderSearch. Finally,
in Preprocessor::HandleIncludeDirective, mirror logic in GCC: the system headerness
of a file being entered can be set due to the #includer or the #includee.
llvm-svn: 56688
directory is shadowed by a user directory in the lookup
path, ignore the user directory not the system one. Not
doing this can affect file lookup and the "is a system
header" bit on locations. For example:
clang -v -I/usr/include inc.c -E | & grep /usr/inc
now prints:
# 1 "/usr/include/i386/_types.h" 1 3 4
# 37 "/usr/include/i386/_types.h" 3 4
# 70 "/usr/include/i386/_types.h" 3 4
instead of:
# 1 "/usr/include/i386/_types.h" 1
# 37 "/usr/include/i386/_types.h"
# 70 "/usr/include/i386/_types.h"
This is part of rdar://6243860.
llvm-svn: 56669
s/ParamAttr/Attribute/g
s/PAList/AttrList/g
s/FnAttributeWithIndex/AttributeWithIndex/g
s/FnAttr/Attribute/g
This sets the stage
- to implement function notes as function attributes and
- to distinguish between function attributes and return value attributes.
llvm-svn: 56623
- Add CodeGenFunction::{EmitReturnOfRValue, EmitIvarOffset}
- Factor EmitLValueForIvar out of EmitObjCIvarRefLValue.
No non-error functionality change (some unsupported errors are
degraded to asserts for the time being).
llvm-svn: 56543
- readonly and readwrite are mutually exclusive.
- assign, copy, and retain are mutually exclusive.
- copy and retain are invalid on non-object types.
- Warn about using default 'assign' property on object types
(attempting to follow gcc behavior).
llvm-svn: 56507
Changed casing of many bug names. The convention will be to have bug names (mostly) lower cased, and categories use some capitalization.
llvm-svn: 56385
GRExprEngine:
Use PostStore in EvalStore.
Use a second version of EvalStore in EvalBinaryOperator to associate the store with the expression on the LHS.
llvm-svn: 56383
For checking if a symbol >= value, we need to check if symbol == value || symbol
> value. When checking symbol > value and we know that symbol != value, the path
is infeasible only if value == maximum integer.
For checking if a symbol <= value, we need to check if symbol == value || symbol
< value. When checking symbol < value and we know that symbol != value, the path
is infeasible only if value == minimum integer.
Updated test case exercising this logic: we only prune paths if the values are
unsigned.
llvm-svn: 56354
'symbol operator-reverse int'. This patch is a combination of code from
Zhongxing Xu and myself (Zhongxing noticed this bug for the cases of
relational operators).
llvm-svn: 56351
"Method accepting NSError** argument should have non-void return value to indicate that an error occurred."
Test case written, but the header needs to be delta-debugged reduced. Will commit shortly.
llvm-svn: 56297
- Enables use of ABIArgInfo::Expand when needed. This greatly
improves our x86 ABI compatibility.
- As the infrastructure for target specific ABI handling isn't built
yet, this change means ABI compatibility on other platforms is once
again broken in a different way than before.
- Upcoming: Figure out how to refactor ABI handling into
targets. More documentation.
llvm-svn: 56293
Ensure that ABIArgInfo::Default implementation for aggregrate types
is sensible (StructRet on return and ByVal on arguments).
No functionality change.
llvm-svn: 56289
motivated because it became clear that the number of subclasses of ProgramPoint
would expand and we ran out of bits to represent a pointer variant. As a plus of
this change, BlockEdge program points can now be represented explicitly without
using a cache of CFGBlock* pairs in CFG.
llvm-svn: 56245
- For investigating warnings in system headers / builtins.
- Currently also enables the behavior that allows silent redefinition
of types in system headers. Conceptually these are separate but I
didn't feel it was worth two options (or changing LangOptions).
llvm-svn: 56163
- Follow C99 behavior of using other operand type when one of
operands is a null pointer constant.
- Fix overenthusiastic devolving of any Objective-C types to id:
o If either operand has an Objective-C object type then:
- If both operands are interfaces and either operand can be
assigned to the other, use that type as the composite type.
- Otherwise, if either type is id, use id as the composite type.
- Otherwise, warn about incompatible types and use id as the
composite type.
- Return handling of qualified idea to separate test following
general pointer type checking.
o Upgraded from old code to allow devolving to id (without warning,
which matches GCC).
- <rdar://problem/6212771>
Add test case for issues fixed above, XFAIL though because it exposed
a new issue in property handling.
llvm-svn: 56135
value).
- Added ABIArgInfo::ByVal (mostly supported) and ABIArgInfo::Expand
(asserted out).
- Added classifyArgumentType which currently just uses
ABIArgInfo::Default or ByVal. This nearly matches old behavior, but
we now set ByVal in a few situations we may have left it off before
(on complex, for example).
llvm-svn: 56086
(1) Additional arguments to variadic methods should have default
promotions applied.
(2) Additional arguments to non-variadic methods were allowed.
llvm-svn: 56084
no method declaration was found.
- This was allowing arrays to pass "by value" among other things.
Add assert in CodeGen that arguments cannot have array type.
llvm-svn: 56080
-Scoping in C99 works good for C++ too, remove the C++-specific comments.
If someone thinks that the C++-specific comments are necessary for clarification, let me know and I'll put them back on.
llvm-svn: 56078
- As a test, enable basic usage for some common x86-32 cases. This
increases our x86-32 compliance (on other targets our compliance
will just be broken in a different way).
llvm-svn: 56051
type of a call.
Change NeXT runtime to use this instead of trying to bitcasting
internally (which doesn't respect the ABI).
Fix subtle bug, use of ConvertTypeRecursive instead of ConvertType is
bad inside GetFunctionType.
llvm-svn: 56050
- Added CodeGenTypes::GetFunctionType, taking a CGFunctionInfo.
- Updated Obj-C runtimes to use this instead of rolling the
llvm::FunctionType by hand.
- Killed CodeGenTypes::{ConvertReturnType, DecodeArgumentTypes}.
Add ABIArgInfo class to encapsulate ABI decision of how to lower types
to LLVM.
- Will move to target sometime soon.
llvm-svn: 56047
- Realized these functions will eventually need access to more data,
moved to CodeGenModule. Eventually they should probably live
together in some other helper class.
llvm-svn: 56039
The 'control' scope is the 'condition' scope of if/switch/while statements and the scope that contains the for-init-statement and 'condition' of a for statement.
e.g:
if (int x = 0 /*'control' scope*/) {
// x will be regarded as part of this substatement scope.
} else {
// and as part of this substatement scope too.
}
llvm-svn: 56020
- Replace string comparisons with pre-defined idents.
- Avoid calling isBuiltinObjCType() to avoid two checks.
- Remove isBuiltinObjCType(), since it was only used in Sema::MergeTypeDefDecl().
- Have Sema::MergeTypeDefDecl() set the new type.
This is a moidified version of an patch by David Chisnall.
llvm-svn: 55990
- Add CodeGenFunction::EmitAnyExprToTemp
o Like EmitAnyExpr, but emits aggregates to a temporary location if
none is available. Seems like this should be simpler (even aside
from using first class aggregates).
- Killed CodeGenFunction::EmitCallArg (just append the pair)
- Conversion of RValues to actual call arguments is now isolated in
CodeGenFunction::EmitCall.
llvm-svn: 55970
aliases.
- Attributes specific to a definition are only set when the
definition is seen.
- Alias generation is delayed until the end of the module; necessary
since the alias may reference forward.
- Fixes: PR2743, <rdr://6140807&6094512>
- Improves: <rdr://6095112> (added XFAIL)
Also, print module on verification failures.
llvm-svn: 55966
- Add CGCall.h for dealing with ABI issues related to calls.
- Add CGFunctionInfo and CGCallInfo for capturing ABI relevant
information about functions and calls.
- Isolate LLVM parameter attribute handling inside CGCall.cpp
llvm-svn: 55963
If you're on some other platform, the correct definition for this macro
would be appreciated; to find the correct definition, just run the
following command:
echo | gcc -dM -E - | grep USER_LABEL_PREFIX
llvm-svn: 55869
This change effects both RecordDecls and CXXRecordDecls, but does not effect EnumDecls (yet).
The motivation of this patch is as follows:
- Capture more source information, necessary for refactoring/rewriting clients.
- Pave the way to resolve ownership issues with RecordDecls with the forthcoming
addition of DeclGroups.
Current caveats:
- Until DeclGroups are in place, we will leak RecordDecls not explicitly
referenced by the AST. For example:
typedef struct { ... } x;
The RecordDecl for the struct will be leaked because the TypedefDecl doesn't
refer to it. This will be solved with DeclGroups.
- This patch also (temporarily) breaks CodeGen. More below.
High-level changes:
- As before, TagType still refers to a TagDecl, but it doesn't own it. When
a struct/union/class is first referenced, a RecordType and RecordDecl are
created for it, and the RecordType refers to that RecordDecl. Later, if
a new RecordDecl is created, the pointer to a RecordDecl in RecordType is
updated to point to the RecordDecl that defines the struct/union/class.
- TagDecl and RecordDecl now how a method 'getDefinition()' to return the
TagDecl*/RecordDecl* that refers to the TagDecl* that defines a particular
enum/struct/class/union. This is useful from going from a RecordDecl* that
defines a forward declaration to the RecordDecl* that provides the actual
definition. Note that this also works for EnumDecls, except that in this case
there is no distinction between forward declarations and definitions (yet).
- Clients should no longer assume that 'isDefinition()' returns true from a
RecordDecl if the corresponding struct/union/class has been defined.
isDefinition() only returns true if a particular RecordDecl is the defining
Decl. Use 'getDefinition()' instead to determine if a struct has been defined.
- The main changes to Sema happen in ActOnTag. To make the changes more
incremental, I split off the processing of enums and structs et al into two
code paths. Enums use the original code path (which is in ActOnTag) and
structs use the ActOnTagStruct. Eventually the two code paths will be merged,
but the idea was to preserve the original logic both for comparison and not to
change the logic for both enums and structs all at once.
- There is NO CHAINING of RecordDecls for the same RecordType. All RecordDecls
that correspond to the same type simply have a pointer to that type. If we
need to figure out what are all the RecordDecls for a given type we can build
a backmap.
- The diff in CXXRecordDecl.[cpp,h] is actually very small; it just mimics the
changes to RecordDecl. For some reason 'svn' marks the entire file as changed.
Why is CodeGen broken:
- Codegen assumes that there is an equivalence between RecordDecl* and
RecordType*. This was true before because we only created one RecordDecl* for
a given RecordType*, but it is no longer true. I believe this shouldn't be too
hard to change, but the patch was big enough as it is.
I have tested this patch on both the clang test suite, and by running the static analyzer over Postgresql and a large Apple-internal project (mix of Objective-C and C).
llvm-svn: 55839
The motivation behind this change is that chaining the RecordDecls is simply unnecessary. Once we create multiple RecordDecls for the same struct/union/class, clients that care about all the declarations of the same struct can build a back map by seeing which Decls refer to the same RecordType.
llvm-svn: 55821
Daniel Dunbar